Simultaneous Liver Kidney Transplantation: A Medical Decision Analysis

Abstract

Background. 

The use of simultaneous liver kidney transplantation has increased dramatically. When the liver and kidney are available from the same deceased donor, what is the best decision? There are two allocation options. In the combined allocation, both organs are allocated to a liver failure (end-stage liver disease [ESLD]) patient on dialysis leaving an end-stage renal disease (ESRD) patient on dialysis. In split allocation, the liver is allocated to the liver failure patient on dialysis and the kidney to the patient with ESRD.

Methods. 

A computerized medical decision analysis was performed using published US survival data. The two options were compared by examining differences in projected quality-adjusted life years (QALYs).

Results. 

Combined allocation was the best strategy (+0.806 QALYs) if liver transplant recipients on dialysis have proportionately worse survival compared with kidney failure alone patients on dialysis. However, because some patients with hepatorenal syndrome recover kidney function post-liver transplant alone (LTA), a second analysis incorporated the possibilities of being dialysis free. If the chance of recovery of renal function is 50% rather than 0%, the decision reversed. Here, the split allocation provided 1.02 more total QALYs than the combined allocation.

Conclusions. 

This study demonstrates that simultaneous liver kidney transplantation is an excellent strategy in most patients with both ESLD and ESRD. However, allocating a kidney to a patient with ESLD, who has the potential to be dialysis free without a kidney transplant, does not maximize overall outcomes when all patients are considered.

One of the most important steps in organ transplantation is a fair and just allocation process. The increase in combined organ transplantation may pose ethical challenges, especially if the conditions of eligibility or the rate of transplantation differ substantially between those needing one or more organs (¹). Since the adoption of the Model for End-Stage Liver Disease (MELD) score for allocating organs in 2002, there has been a steady increase in simultaneous liver kidney transplantation (SLK) (^1–3). A recent consensus conference reported their indications for SLK (³). In addition to recommending SLK in patients with end-stage liver disease (ESLD) on dialysis for chronic kidney disease (CKD), they also recommend that ESLD patients with prolonged acute kidney injury (AKI) or chronically low glomerular filtration rate (GFR) less than 30 mL/min be eligible for both organs (³). These latter recommendations could potentially increase SLK further (¹).

Despite the enthusiasm of SLK, a recent analysis by Locke et al. (⁴) provides evidence that increases in SLK have not been associated with improved overall short-term outcomes. Mortality at 1 year presently exceeds 20%, such that many kidney allografts are lost with the patient (^{2, 3}). In comparison, combined patient and graft loss within the first year is less than 10% for kidney transplant alone recipients. Intuitively, it makes sense to benefit two people rather than give both organs to one person. The preferential use of kidneys in liver recipients will increase wait times for other patients with end-stage renal disease (ESRD). There are a wide range of SLK rates in individual centers, which might suggest different indications at different centers (³). In one center, approximately 50% of all the liver transplants performed were SLK (³).

In contrast to the aforementioned data, there are several reasons to support preferential kidney allocation to SLK recipients. Patient survival is better at the end of the first year with combined rather than liver transplantation alone (LTA) (^{3, 5}). There is evidence that the kidney allograft lasts longer in liver transplant recipients from the same donor (^{6, 7}). There is only one operation. Also, dialysis seems to be tolerated poorly in liver transplant recipients compared with matched kidney failure only dialysis patients (⁸). These observations support the contention that the kidney “follows” the liver into recipients who can benefit from both.

If there is a kidney and a liver available from the same donor, should both go to a patient needing both leaving a patient with ESRD on dialysis or should a potential kidney recipient get the kidney and the patient with uremic ESLD get the liver alone? Because there are no randomized trials, the purpose of this medical decision analysis was to quantify the outcomes of these two allocation options given the available literature. Because the recent consensus conference also recommended that ESLD patients with prolonged AKI or CKD (not on dialysis) also be given both organs, a second analysis was performed in this population (³).

RESULTS

Figure 1 shows the medical decision tree, and Table 1 shows the probabilities used in the model. The primary analysis, where dialysis (ESRD) was assumed to be certain in patients with ESLD, showed that combined allocation provided 0.806 more quality-adjusted life years (QALYs) than split allocation (Table 2). The kidney organ was projected to provide more QALYS when transplanted into the SLK recipient (3.626 QALYs) than to the kidney transplant alone recipient (2.819 QALYs).

FIGURE 1.: Medical decision tree. All liver alone transplant recipients in split allocation remain on dialysis. In second analysis, there was a probability that some of the liver transplant alone patients were dialysis free posttransplant but later could develop end-stage renal disease. SLK, simultaneous liver kidney transplant; LTx, liver transplant; KTx, kidney transplant; LTA, liver transplant alone.

TABLE 1: Model variables, ranges, and description

TABLE 2: QALY outcomes for the analyses and options

In a threshold analysis, if the 1-year mortality was very high (>39% for SLK) for the liver transplant recipient, the split option was preferred (Fig. 2a). The combined option provided more QALYS as long as the relative risk of death for liver transplant patients on dialysis was more than 1.8 (baseline 3.0; Fig. 2b). In a series of one-way sensitivity analyses, all other variables did not change the conclusions of this analysis over their range of values (Figure 3A).

FIGURE 2.: (a) Sensitivity analysis: one-way sensitivity analysis for mortality at 1 year in liver transplant recipients on average benefits for combined and split allocations. The benefit diminishes as the overall mortality increases and favors split allocation if simultaneous liver kidney transplant 1-year mortality rates are more than 39% (model assumption 22%). Expected values in quality-adjusted life years (QALYs). (b) Sensitivity analysis: One-way sensitivity analysis for increase in relative risk of death for a liver transplant patient on for combined and split allocations. The relative benefit increases for the combined option as the relative risk of death exceeds 1.8 (model assumption 3.0). Expected values in QALYs. (c) Sensitivity analysis: two-way sensitivity analysis (QALYS) comparing variation in the proportion that is dialysis free with the subsequent rate of later dialysis. The greater the proportion that are dialysis free initially and a lower rate of later dialysis favor split allocation. ESRD, end-stage renal disease; RR, relative risk.

FIGURE 3.: (a) One-way sensitivity analyses for the primary analysis (end-stage renal disease [ESRD] certain in end-stage liver disease [ESLD] patients). (b) One-way sensitivity analysis for the second analysis (ESRD uncertain in ESLD patients). ESRD, end-stage renal disease; QALYs, quality-adjusted life years; SLK, simultaneous liver kidney transplant; LTA, liver transplant alone; RR, relative risk.

In the second analysis, where the need for dialysis was not certain in the liver transplant recipients, the split option provided 1.02 more QALYs than the combined option (Table 2). In this analysis, we assumed that the probability of being dialysis free was 50% but that the need for dialysis at a later time point was 10% per year. The conclusions of the analysis did not change when the other variables were examined in a sensitivity analysis (Fig. 3b). In a threshold analysis, only when the relative risk of death for the liver failure patient on dialysis was more than 7-fold, the combined option was preferred. Figure 2(c) shows a two-way sensitivity analysis for both freedom of dialysis initially posttransplant and the need for dialysis at a later date. One example where the combined option would provide more net benefit than the split option would be if the chance of being dialysis free was less than 10% (the alternative explanation is that there would be a 90% chance that the patient required permanent dialysis post-LTA with a low subsequent probability of needing dialysis in the future). A second example supporting the combined option would be if the patient with ESLD had a 60% chance of not needing dialysis initially but had a risk of subsequent need for permanent dialysis of 40% per year.

DISCUSSION

This is the first medical decision analysis to examine the net benefit of a kidney transplanted into an SLK recipient when the kidney and liver are available from the same deceased donor. This analysis quantifies the arguments for the use of a kidney in SLK and provides threshold values for variables where the maximizing the net benefits of the kidney change from being used in the patient with ESLD to being transplanted elsewhere. Despite the lack of randomized control trial data, the use of SLK seems appropriate in many patients who are on the liver transplant wait list and who have established ESRD. In this circumstance, the relatively high early transplant-related mortality associated with liver transplantation is offset by the higher mortality rate of liver transplant recipients on dialysis, creating greater net benefit for the SLK recipient. This may be counterintuitive to some clinicians who might observe the relatively high early mortality of SLK compared with kidney alone transplantation in patients with ESRD as being prohibitive (mortality rates in this latter group are mostly less than 6% after the first-year postkidney transplantation).

This decision analysis shows that the argument most supportive of SLK is the high relative mortality of liver failure patients with ESRD on dialysis. The arguments of better early patient survival (compared with LTA) and better long-term kidney graft function are quantitatively less important. In the study by Al Riyami et al. (⁸), those liver transplant recipients who later required dialysis had poor outcomes when compared with others in the Canadian Organ Replacement Registry. Because they show that liver transplant patients proceeding to a kidney transplant did as well as matched kidney transplant recipients, the presumed net benefit of a kidney transplant compared with being on dialysis will be greater for a liver transplant recipient than a general dialysis person (⁸). In a preliminary examination of the US database, the relative risk was much higher. Liver transplant recipients awaiting a kidney transplant had a waitlist mortality of 25.2 per 100 patient-years compared with a mortality rate of 8.9 for ESRD awaiting a kidney alone (⁹). The reasons for this excessively high mortality deserve further study.

Although the overall findings of this analysis support the consensus guidelines, there are examples where the net benefit of the kidney may be less in the SLK recipient compared with elsewhere. If the early mortality rate postliver transplantation is exceptionally high (>40%), as might be observed in retransplantation for hepatitis C recurrence, the net benefit of the kidney is greatly diminished (Fig. 2a) (¹⁰). In addition, the analysis also does not support using SLK in patients with AKI or CKD not on dialysis (ESRD uncertain). In the circumstances described earlier, SLK will provide better outcomes than LTA alone (see Table 2, QALYs for SLK greater than LTA) because the risk of ESRD is always higher in the LTA recipient and the model assumes that there is no added penalty to perform a kidney transplant with a liver. However, SLK does not result in the best overall (societal) outcomes when patients with ESRD left on dialysis are considered. In these circumstances, a significant proportion of the patients will be given a kidney that was not required and will deprive other dialysis patients in need. Given the wide variation in SLK rates, it is likely that SLK is underused in some and overused in other transplant centers. This analysis should stimulate centers to reevaluate their practice in light of this quantitative analysis.

The consensus guidelines currently recommend SLK in patients with AKI including hepatorenal syndrome on dialysis more than or equal to 8 weeks and a serum creatinine more than 2.0 mg/dL (⁹). The literature on the likelihood of renal recovery from AKI such as hepatorenal syndrome is mixed. An early study found that 90% of patients with hepatorenal syndrome recover kidney function posttransplant, whereas a more recent study reported that the number of patients developing ESRD is 25% with a poor ability to predict recovery based on clinical variables (^{11, 12}). It is not clear to what extent the requirement of 8 weeks of dialysis increases the positive predictive value for developing ESRD, because recovery of renal function has been reported to occur in a liver transplant recipient on dialysis for 8 months (¹³). This medical decision shows that the positive predictive value should be very high for preferential allocation to SLK in this cohort based on maximizing net benefits.

The consensus guidelines also currently recommend SLK for a GFR less than or equal to 30 mL/min with evidence of CKD or CKD with a biopsy demonstrating more than 30% glomerulosclerosis or fibrosis (⁹). In a recent study by O'Riordan et al., a risk score was developed to predict a low GFR (<30 mL/min) at 1 year post-LTA. However, the numbers progressing to this level of function were small, and the score was also not able to predict the development of ESRD because this event was also low (¹⁴). In a study by Bahirwani et al. (¹⁵), patients with increased serum creatinine for more than 12 weeks preliver transplant had a high risk (hazard ratio 5.3) of developing an estimated GFR less than 20 mL/min, but only 25% did so during a 3-year period. In addition, five of the eight patients with diabetes mellitus progressed, suggesting that the type of CKD may be an important predictor of progression. In our second analysis, the rate of developing ESRD was 10% per year similar to rate observed by Bahirwani et al. (¹⁵). However, this medical decision analysis would not support allocating both organs to the same recipient for an ESRD rate of 10% per year when considering total net benefits of all parties. Although kidney biopsy has been recommended with supportive evidence from observational series, this strategy remains untested and questionable because the predictive accuracy for histology (including degree of glomerulosclerosis) in kidney transplantation is poor (^{1, 16, 17}). More accurate assessments of kidney function in this cohort, the trajectory of kidney function change, and kidney histologic analysis (degree of damage and specific lesion) might better predict who needs an SLK. The systematic collection of these data in a large LTA cohort will be a challenge but necessary to develop an accurate prediction score. Follow-up measurements of residual native kidney function in SLK patients might also be informative (¹⁸).

There are limitations to this model. At first glance, the model is an inadequate description of clinical practice. One could argue that the patient with ESRD in the combined allocation option will have an opportunity to receive a kidney transplant within a short period of time and that the analysis should include this probability as it would strengthen the case for SLK. Because this requires equal treatment to both allocation arms, the analysis must also allocate a kidney to the LTA in the split allocation. In its simplicity, if there were one liver and two kidneys being offered, there would be no need for this medical decision analysis. The important comparison is the second analysis when ESRD is uncertain in the patient with ESLD. If this recipient with ESLD is given a kidney that was not necessary, then there is a potential lost benefit to a patient with ESRD who remains on dialysis. The calculation of this lost benefit uses the same rationale and methodology as others have used to calculate the impact of a lost organ donor. Because there is always someone dying on the wait list, the appropriate calculation is to examine the impact of this organ over the life expectancy of the graft (¹⁹). The longer the patient with ESLD has a kidney transplant that was not required, the greater the lost benefit to a patient with ESRD remaining on dialysis. This analysis calculates that lost benefit.

One of the most important variables affecting the conclusions of this analysis is the relative patient survival. Several studies show that the net benefit of kidney and liver transplantation vary greatly for individual wait list recipients (^{20, 21}). It is conceivable that many actual transplant pairs would have outcomes different from this model; however, the analysis of patient survival probabilities within their 95% confidence intervals (CIs; Fig. 3) had little effect on the conclusions even when examining multiple changes (data not shown). Rather than be a definitive analysis for all patient subsets (age, MELD score, donor quality, etc.), this model is a framework for further analysis and discussion. The use of a net benefits scheme to allocate kidney organs has been in development since 2004; however, this will not be implemented for many reasons (²²). This move away from utility-based arguments toward equity may at first seem to undermine this analysis. However, we believe that our analysis remains meaningful. The consensus conference argues that some ESLD patients with AKI or CKD (<30 mL/min/1.73 m²) not yet on dialysis receive a SLK (³). Not only does this result in significantly less overall net benefits but also this recommendation may be inequitable if other patients on the list for a kidney transplant must meet stricter criteria (irreversible, advanced, and progressive kidney disease) and wait longer.

In summary, this study suggests that SLK is an excellent strategy in patients with both ESLD and irreversible kidney failure. The study also suggests that split allocation is a better option when the patient with ESLD has reversible kidney failure.

From this analysis, we also propose the following:

1. More research is needed to understand the increased relative risk of mortality for liver transplant recipients on dialysis compared with the general sample of patients with ESRD. This is the most important factor that counter balances the high early mortality (with kidney graft loss) in SLK when calculating the net benefit of a kidney transplant.
2. SLK is reasonable in acceptable liver transplant candidates with ESRD if overall patient mortality is not excessive. Although the MELD score predicts death on the wait list, predictors of excessive death postliver transplant should be considered before automatically assuming that a kidney should also be transplanted along with the liver.
3. ESLD patients with prolonged AKI and low GFR CKD should not automatically receive a kidney transplant at this time. Based on this analysis, prediction equations should be developed with prospective clinical experience to determine who has a very low probability of recovery of kidney failure or a very high probability of developing ESRD post-LTA. The current recommendations of the consensus conference do not have sufficient evidence to support their implementation at this time (³).

The nephrology community must work together with our liver colleagues at providing the best care to all patients with kidney disease. This is an excellent area for future research.

MATERIALS AND METHODS

The medical decision analysis was performed using published guidelines (²³). Two allocation options were compared for a kidney and liver from the same deceased donor. The combined allocation option was to leave the patient with ESRD on dialysis and perform an SLK in one recipient requiring both, and the split allocation option was to perform a kidney transplant in the patient with ESRD and perform a liver transplant alone (LTA) in the patient with uremic ESLD. The incremental number of QALYs in the split versus the combined option was the main outcome. Because the patient with ESLD always receives a liver transplant, the analysis can be conceived as determining who derives the most benefit from the kidney transplant.

Probabilities for patient and graft survival were taken from the published recent registry analyses and are defined in Table 1 (^24–26). Quality of life scaling factors for dialysis, functioning kidney transplant, and liver transplant were taken from a prior economic evaluation (¹⁹). No specific quality of life scaling factor was assigned for an ESLD patient with hepatic decompensation because in both options, the liver failure patient received a liver. The model is shown in Figure 1.

ESRD Patient

In the primary analysis, the patient with ESRD was presumed to be on dialysis. In the combined allocation option, the patient remained on dialysis throughout. The dialysis mortality rate was set at the overall wait list mortality rate (²⁴).

In the split allocation option, the patient with ESRD received a kidney transplant. Patients surviving a failed kidney transplant returned to dialysis at an annual higher mortality rate (²⁴).

ESLD Patient ESRD Certain

In the primary analysis, the patient with ESLD was assumed to be on dialysis requiring from chronic irreversible disease. In the combined allocation option, this recipient received both organs. We assumed the following:

• Long-term patient mortality rates (after the first year) in those with functioning kidney and liver transplants would be similar to the general liver transplant alone recipient population.
• Patients with a failed liver graft could be retransplanted if needed.
• Kidney graft failure rates in SLK patients would be proportionately less if both organs are transplanted from the same donor (^{6, 7}).
• SLK patients who lost the kidney allograft and required dialysis had proportionately higher mortality rates (^{8, 9}). We estimated the relative risk from the study by Al Riyami et al. (⁸) to be 1.5. The preliminary report from the US registry analysis provided a hazard ratio of 3.03 (95% CI 2.78–3.27) when comparing the risk of death on the wait list for a dialysis-dependent liver transplant patient compared with a wait-listed ESRD patient (⁹). Given the discrepancy in these reports, we used the relative risk of 3.0 but examined a wide range of values (1.0–6.0) in a threshold analysis. All SLK patients who suffered kidney graft loss were assumed to remain on dialysis.

In the split allocation model, the patient with ESLD on dialysis received an LTA. We assumed the following:

• Patient survival at the end of the first year was 5.7% lower in these recipients compared with SLK recipients (^{3, 5}).
• Mortality rates after the first year were calculated to be the same as a wait-listed ESRD patient multiplied by the increase relative risk estimated (as earlier) for liver transplant recipients on dialysis (⁹). The model assumed that the patient remained on dialysis throughout.

ESLD Recipient ESRD Uncertain

In the second analysis, split allocation allowed for the possibility that patients with ESLD and AKI might recover kidney function when given an LTA. Similarly, ESLD patients with CKD not yet on dialysis might be given an LTA and subsequently develop ESRD later posttransplant. In both cases, after an LTA, patients might be free of dialysis. Because this freedom may be short lived, we incorporated rates of ESRD for the LTA recipients. Because recent evidence does not support increased early mortality in non-ESRD patients receiving an LTA, 1-year mortality was assumed equivalent for SLK and LTA recipients in this second analysis (⁵).

The analyses were performed on TREEAGE PRO 2004 (TreeAge Software, Inc., Williamstown, MA). Multiple sensitivity analyses were performed over the range of the variables shown in Table 1. If available, the 95% CIs were used to establish variable ranges. Thresholds for the changes in the direction of the conclusion were calculated for key variables. A discount rate (rate of time preference) of 3% was used on health outcomes.